particle motion, coating and drying in wurster fluidized...

THESIS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

Particle motion, coating and drying

in Wurster fluidized beds

- An experimental and discrete element modeling study

LIANG LI

Chemical Engineering

Department of Chemistry and Chemical Engineering

CHALMERS UNIVERSITY OF TECHNOLOGY

Gothenburg, Sweden 2015

Particle motion, coating and drying in Wurster fluidized beds - An experimental and

discrete element modeling study

LIANG LI

ISBN: 978-91-7597-251-0

© LIANG LI, 2015

Doktorsavhandlingar vid Chalmers tekniska högskola

Serie Nr 3932

ISSN 0346-718X


Chalmers University of Technology

SE-412 96 Gothenburg, Sweden

Telephone +46(0)31-772 1000

Printed by Chalmers Reproservice

Gothenburg, Sweden 2015

Cover: The cover image shows a DEM-CFD simulation of particle motion in the LAB CC system.

The particles are colored by their vertical velocity, and the central plane is colored by air velocity.

The red color represents a high value, and the blue color represents a low value. The shadow

illustrates the walls of the fluidized bed and the Wurster tube, and the grid used.

i

Particle motion, coating and drying in Wurster

fluidized beds - An experimental and discrete

element modeling study

Liang Li


Chalmers University of Technology

SE-412 96 Gothenburg, Sweden

Abstract

This thesis focuses on developing a mathematical model that is capable of predicting the fluidized

bed coating process. First a basic understanding of pellet motion in a Wurster fluidized bed was

established by conducting a series of experiments using the positron emission particle tracking

(PEPT) technique. The PEPT results, such as the particle velocity, the cycle time distribution

(CTD) and the residence time distribution (RTD) of particles in different regions of the fluidized

bed, were selected to evaluate the model for pellet motion based on the discrete element method,

computational fluid dynamics (DEM-CFD). In an effort to refine the sub-models involved in the

DEM, the effects of the drag model were investigated. With the validated DEM-CFD model, the

detailed pellet motion in the spray zone of the fluidized bed was studied. In order to identify the

underpinning mechanisms by which coating thickness changes, pellets of different sizes were

employed, and a simple model for predicting the growth of pellets was developed based on data

available from the DEM-CFD results and the PEPT experiments. This predictive model was then

evaluated experimentally. In addition, a model for drying of pellets was developed.

In the PEPT experiments, it was found that, for the parameters studied, particles spend

approximately 12–29% of the cycle time in the Wurster tube. It was observed that particles tend

to recirculate in the Wurster tube and sneak out from below the tube. In comparison with the

PEPT experiments, the coupled DEM-CFD simulations showed close agreement with respect to

the CTD and the RTD of particles in different regions. Using the validated DEM-CFD model,

large particles were found to spend longer time in the spray zone and move closer to the spray

nozzle. The latter effect provides evidence that large particles can shield small particles from

spray droplets. Both of these effects suggest that large particles receive a greater amount of

coating solution per particle cycle. A simple conceptual model was then developed to predict the

effects of the residence time of particles in the spray zone, the particle cycle time, and the

entrance distance on the relative rate of the increase in film thickness between large and small

particles. By comparing predicted and measured relative rates of increase in coating thickness

between large and small particles, it was confirmed that large particles grow faster than small

particles.

Keywords: Fluidized bed, particle, coating, drying, airflow, DEM, CFD, PEPT, Wurster, spray

iii

List of publications

The thesis is based on the following publications:

Paper I:

Li, L., Rasmuson, A., Ingram, A., Johansson, M., Remmelgas, J., von Corswant, C., & Folestad, S.

(2015). PEPT study of particle cycle and residence time distributions in a Wurster fluid

bed. AIChE Journal, 61(3), 756-768.

Paper II:

Li, L., Remmelgas, J., van Wachem, B. G. M., von Corswant, C., Johansson, M., Folestad, S., &

Rasmuson, A. (2015). Residence time distributions of different size particles in the spray

zone of a Wurster fluid bed studied using DEM-CFD. Powder Technology, 280(0), 124-134.

Paper III:

Li, L., Remmelgas, J., van Wachem, B. G. M., von Corswant, C., Folestad, S., Johansson, M., &

Rasmuson, A. (2016). Effect of Drag Models on Residence Time Distributions of Particles

in a Wurster Fluidized Bed: a DEM-CFD Study. KONA Powder and Particle Journal,

10.14356/kona.2016008.

Paper IV:

Li, L., Hjärtstam J., Remmelgas, J., von Corswant, C., Folestad, S., Johansson, M. & Rasmuson,

A. (2015). Growth of differently sized particles in a fluidized bed coater. Submitted to

Powder Technology.

Paper V:

Li, L., Remmelgas, J., van Wachem, B. G. M., Hjärtstam J., von Corswant, C., Folestad, S. &

Rasmuson, A. (2015). A DEM-CFD study on drying of coated particles in a fluidized bed,

manuscript in preparation.

iv

Contribution report

In addition to being the lead author in the above publications, the candidate has contributed in

the following aspects to each paper below:

Paper I:

Prepared the experiments, analyzed the experimental data, and together with my co-authors,

interpreted the results.

Paper II and III:

Performed the simulations, developed the code for pre- and post-processing, and together with

my co-authors, interpreted the results.

Paper IV:

Performed the experiments with Hjärtstam J., analyzed the experimental data, and together

with my co-authors, interpreted the results.

Paper V:

Performed the experiments with Hjärtstam J., carried out simulations and the analysis,

developed the code for pre- and post-processing, and together with my co-authors, interpreted the

results.

v

Acknowledgements

This study was undertaken with financial support from the EU Seventh Framework Programme,

AstraZeneca and Chalmers University of Technology.

First I would like to thank Anders Rasmuson and Staffan Folestad for giving me this unique

opportunity to be involved in this exciting project. Thanks Anders for always being available and

patient, for your motivation and guidance and for your encouragement. I would also like to thank

Staffan for your broad view, knowledge and nice talks not only on the project but also on a

variety of aspects.

Johan Remmelgas, as my co-supervisor and mentor, spent plenty of time and energy in helping

me with problem solving, developing codes, examining and discussing results, writing and

rewriting, as well as English. I appreciate all this very much. Many thanks to you.

To my co-supervisor Christian von Corswant, I am grateful for your valuable comments and for

the insightful discussions we had.

I am also thankful to Berend van Wachem at Imperial College London for introducing me to

MultiFlow and for remote but rewarding guidance and discussions.

I want to acknowledge Mats Johansson and Johan Hjärtstam for nice cooperation on the

experimental work. In the experiments, I also received help from Bindhu Gururajan, Andy

Ingram and Sven Karlsson, Anders Holmgren, Brita Sjöblom, Hanna Matic, Magnus Fransson,

Livia Cerullo, Alvaro Diaz-Bolado; thank you all. During simulations, I received a lot of support

from Mikael Bengtsson, Andreas Loong, Martin Budsjö and Nima Namaki at the high-

performance computing (HPC) team, AstraZenceca; thank you very much.

To Lilia Ahrné, thank you for coordinating this beautiful network and organizing many

wonderful events. To PowTech members, thank you very much for sharing during these years

and it was really nice meeting you. In particular, it was great to have Duy Nguyen, Loredana

Malafronte, Pooja Shenoy and Epameinondas Xanthakis in Gothenburg; I have fond memories of

having beers, chats and dancing and traveling with you and David Valdesueiro.

I also enjoyed being the roommate of Kurnia Wijayanti; thank you for sharing relevant and

irrelevant stories. Of course I am pleased to have friendly and helpful colleagues at Chalmers

and AstraZeneca, who created an enjoyable working environment.

My special thanks go to my big brother Chaoquan Hu, Ms. Yan and their lovely son, Chengsong.

May you succeed in achieving great happiness in the future!

Last but not least, I devote my deepest words to my beloved family and An’an: I love you!

“爸妈，晶晶，我爱你们，谢谢你们。因为有你们，让我懂得生命中什么是重要；因为有你们，让我有

勇气面对一切；因为有你们，一切才变得有意义！你们都要好好的！”

https://www.google.se/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&ved=0CC0QFjAA&url=https%3A%2F%2Fwww.chalmers.se%2Fchem%2Fsumo-en%2Fpeople%2Fhanna-matic&ei=dmDdUsKWNcy_ywOrx4AI&usg=AFQjCNEu5LifvQ3gQB1x75nt7yerOF5log&sig2=iw0Z0Q8g4_lOte0pKDlwug&bvm=bv.59568121,d.bGQ

vii

Contents

1. Introduction ................................................................................. 1

1.1. Background ..................................................................................................... 1

1.2. The Wurster coating process .......................................................................... 3

1.3. Objectives ........................................................................................................ 5

1.4. Outline of the thesis ........................................................................................ 6

2. Set-up of the system ..................................................................... 7

2.1. STREA-1™ ...................................................................................................... 7

2.2. LAB CC ............................................................................................................ 9

3. Modeling methods ...................................................................... 11

3.1. Overview of multiphase models .................................................................... 11

3.1.1. Eulerian-Eulerian approach ................................................................... 11

3.1.2. Eulerian-Lagrangian approach ............................................................... 11

3.2. Movement of pellets ...................................................................................... 12

3.2.1. Discrete element method ......................................................................... 13

3.2.2. Gas flow .................................................................................................... 16

3.2.3. Drag model ............................................................................................... 16

3.3. Numerical methods ....................................................................................... 19

3.4. Drying of pellets ............................................................................................ 20

4. Experimental .............................................................................. 25

4.1. PEPT .............................................................................................................. 25

4.1.1. The PEPT measurement system............................................................. 25

4.1.2. Materials .................................................................................................. 26

4.1.3. The tracer particle ................................................................................... 26

4.1.4. Operating parameters ............................................................................. 26

viii

4.1.5. Post-processing ........................................................................................ 27

4.2. Coating experiments ..................................................................................... 30

4.2.1. Materials .................................................................................................. 30

4.2.2. QicPic analysis ......................................................................................... 30

4.2.3. Coating solution ....................................................................................... 31

4.2.4. Determination of growth of pellets ......................................................... 31

4.3. Drying experiments....................................................................................... 32

4.3.1. Method ..................................................................................................... 32

4.3.2. Coating solution ....................................................................................... 32

4.3.3. Operating parameters ............................................................................. 32

5. Results and discussion .............................................................. 35

5.1. Characteristics of particle motion (Paper I) ................................................. 35

5.2. Validation and development of DEM-CFD model (Paper II) ...................... 39

5.2.1. Calibration of inlet flow ........................................................................... 39

5.2.2. Model validation ...................................................................................... 40

5.2.3. Evaluation of drag model (Paper III) ...................................................... 43

5.3. A predictive model for growth of pellets (Paper II and Paper IV) .............. 44

5.4. Particle drying ............................................................................................... 50

6. Conclusions and outlook .......................................................... 51

References ....................................................................................... 55

Appendix: Flow distribution at the distributor ......................... 59

ix

Nomenclature

Abbreviations

CFD computational fluid dynamics

CTD cycle time distribution

DEM discrete element method

EC ethyl cellulose

EE Eulerian-Eulerian

EL Eulerian-Lagrangian

HPC hydroxypropyl cellulose

IBM immersed boundary method

FVM finite volume method

LBM Lattice-Boltzmann method

LES large eddy simulation

MCC microcrystalline cellulose

PEPT positron emission particle tracking

PSD particle size distribution

RHS right hand side

RSD relative standard deviation

RTD residence time distribution

SEM scanning electron microscope

TFM two-fluid model

VMD volume mean diameter

Roman letters

A area

B inertial resistance coefficient

x

𝑐𝑝 specific heat capacity

𝐶 discharge coefficient

𝐶𝐷 drag coefficient

𝑑 diameter of the particle

D pressure loss coefficient

𝐷𝐴𝐵 diffusivity of the solvent in air

𝐷𝑠 aperture width

𝐸 Young’s modulus

𝑭𝒄 contact force

𝒈 gravity

ℎ convective heat transfer coefficient

𝐻 shear modulus

𝐻𝑒𝑣𝑝 latent heat of evaporation

𝐼 the moment of inertia

𝑘 spring stiffness

𝑘𝐵 Boltzmann’s constant

𝑘𝑝𝑔 mass transfer coefficient

𝑙𝑓𝑖𝑙𝑚 thickness of the coating film

𝑙𝑝 thickness of the porous zone

𝐿 height of the spray zone

𝑚𝑝 particle mass

𝑚𝑝∗ effective particle mass

�̇� spray rate

𝑀 molar weight

𝑁𝑐𝑦𝑐𝑙𝑒 number of coating cycles

𝑁𝑖 degree of the polymerization, 𝑖 = 2

𝑁𝑅𝑒 screen Reynolds number

𝑁𝑤,𝑝 mass flux of liquid through the surface of the pellet

𝑃 pressure

PD pressure drop through the wire mesh screen

xi

𝑟 radius of the particle

𝑟𝑒𝑛𝑡𝑟𝑎𝑛𝑐𝑒 entrance distance

𝑆𝑐 Schmidt number

𝑆ℎ Sherwood number

𝑅 gas constant

𝑅𝑒 Reynolds number

𝑡 time

𝑡𝑐𝑦𝑐𝑙𝑒 cycle time

𝑡𝑠𝑝𝑟𝑎𝑦 residence time in the spray zone

𝑇 temperature

𝑻 torque

𝒖 fluid velocity

𝒗 particle velocity

𝑉 volume

𝑥 mole fraction of component

𝑌 mass fraction

Greek letters

𝛼 constant related to the coefficient of restitution

𝛽 interphase momentum transfer coefficient

휀 volume fraction or porosity

𝛿 displacement of the particle

𝜂 damping coefficient

𝜃 spray half-angle

𝜆 thermal conductivity

𝜇 friction coefficient

𝜌 density

𝜎 Poisson’s ratio

𝝉 viscous stress tensor

xii

𝜙 volume fraction of competent

𝝎 rotational velocity

Subscripts

𝑔 gas phase

𝑖 the ith particle

𝑗 the jth particle

𝑛 normal direction

𝑝 particle

𝑠𝑢𝑏 polymer substance

𝑡 tangential direction

𝑤 solvent

1, 2 component 𝑖

xiii

List of figures

Figure 1.1 Layout of a pellet and its usage, partly revised from the work of Kearney

and Mooney (2013)........................................................................................... 1

Figure 1.2 Schematic of different drug release profiles ................................................... 2

Figure 1.3 Schematic of the Wurster process and the different regions: 1) the spray

zone, 2) the Wurster tube, 3) the fountain region, 4) the downbed region,

and 5) the horizontal transport region. ........................................................... 3

Figure 2.1 The dimensions of the Wurster bed, a) side view showing the different

regions: (1) the Wurster tube, (2) the fountain region, (3) the downbed

region, and (4) the horizontal transport region; b) top view. .......................... 8

Figure 3.1 A schematic of the soft sphere model ............................................................ 14

Figure 3.2 Comparison of drag models, the relative velocity between gas flow and

particles is (a) 5 m/s and (b) 1 m/s, respectively. .......................................... 18

Figure 3.3 The relation between relative humidity and moisture content, e.g. for

HPMC. ........................................................................................................... 22

Figure 4.1 Schematic of the PEPT measurement system .............................................. 25

Figure 4.2 An example showing the axial (y) coordinates of the tracer particle

positioned on top of the nozzle with varying errors. ..................................... 28

Figure 4.3 The effect of the mean sampling interval on the mean cycle time (partition

gap 10 mm, batch size 200 g, particle VMD 1749 μm, tube length 150 mm,

fluidization airflow rate 73.3 m3/h, and atomization airflow rate 3.50 m3/h).

....................................................................................................................... 29

Figure 5.1 Different types of trajectories: (a) cycle, and (b) recirculation. The bold lines

are the walls of the fluidized bed and the Wurster tube. ............................. 35

Figure 5.2 Residence time distributions in specific regions, with mean values of all

cycles and ideal cycles provided (partition gap 15 mm, batch size 200 g,

particle VMD 1749 μm, tube length 150 mm, fluidization airflow rate 73.3

m3/h, and atomization airflow rate 3.50 m3/h), the bin size for the fountain

and the downbed regions: 0.1 s, for the Wurster tube: 0.5 s, and for the

horizontal transport region: 1 s. .................................................................... 36

Figure 5.3 An example of the axial (above) and radial (below) position of a tracer

particle during 70 s. ....................................................................................... 37

xiv

Figure 5.4 The particle velocity field at a vertical cross-section through the center of

the bed (a) simulated using DEM-CFD and (b) measured using PEPT (VMD

1749 μm, batch size 200 g, fluidization airflow rate 73.3 m3/h and

atomization airflow rate 3.50 m3/h). ............................................................. 40

Figure 5.5 The cycle time distribution (for cycles shorter than 25 s) calculated in the

DEM-CFD simulation and measured in the PEPT experiment (VMD



Figure 5.6 The mean residence times of particles in different regions (VMD 1749 μm,

batch size 200 g, fluidization airflow rate 73.3 m3/h and atomization airflow

rate 3.50 m3/h). .............................................................................................. 41

Figure 5.7 The mean cycle time and RSD for different batch sizes (VMD 1749 μm,

fluidization airflow rate 73.3 m3/h and atomization airflow rate 3.50 m3/h).

....................................................................................................................... 42

Figure 5.8 The cycle time distribution of particle cycles shorter than 25 s (VMD



Figure 5.9 The residence times of particles in different regions for the base case (VMD



Figure 5.10 A schematic of the spray zone. ...................................................................... 45

Figure 5.11 The measured cycle time of small and large particles in the spray zone for

different mixtures of small and large particles. ........................................... 47

Figure 5.12 The simulated residence time of small and large particles in the spray zone

for different mixtures of small and large particles. ...................................... 47

Figure 5.13 The simulated entrance distance into the spray zone of small and large

particles for different mixtures of small and large particles. ....................... 48

Figure 5.14 The predicted relative rate of increase in the coating thickness between the

large and small particles for different mixtures of small and large particles.

....................................................................................................................... 48

Figure A.1 A 1/6th model of the Strea-1 fluidized bed, including the wire mesh screen as

a porous zone. ................................................................................................ 59

Figure A.2 The percentage of total airflow passing through the central base of the

distributor after the porous zone, i.e. the wire mesh screen, with a spacing

of 3 mm and 5 mm. ........................................................................................ 61

xv

List of tables

Table 2.1 The dimensions of the STREA-1 and LAB CC fluidized bed (in mm). ........... 8

Table 4.1 The configuration and operating parameters for each PEPT run. ............... 27

Table 4.2 Specification of uncoated small and large particles for different mixtures. 30

Table 4.3 The operating parameters used in the drying experiments. ........................ 33

Table 5.1 The size of the coated small and large particles for different mixtures. ...... 49

1. Introduction

1.1. Background

Particle coating is a unit operation that is used in a variety of industries. For

example, in the food industry, ingredients are coated to mask flavor or to

improve stability and shelf life (Teunou & Poncelet, 2002). In the agricultural

industry, seeds are coated to make them easier to handle, to allow them to be

planted by machine, and to improve germination and crop yields (Filho et al.,

1998).

Figure 1.1 Layout of a pellet and its usage, partly revised from the work of Kearney

and Mooney (2013).

In the pharmaceutical industry, which is the focus of this thesis, pellets are

coated with drug substances and functional films, for example films for

1. Introduction 2

controlled release (Jono et al., 2000; Chopra et al., 2002). In its simplest form,

such a controlled-release pellet consists of, from inside to outside, a core pellet, a

drug layer, and a controlled release film, as sketched in Figure 1.1. The core

pellet can be made of sand, sugar, or cellulose while the controlled release film

typically consists of a polymer or polymer blend that has been carefully selected

to provide the desired drug release rate. When the patient swallows the medicine,

the drug substance is released in the gastrointestinal tract and ultimately taken

up in the blood. Preferably, the drug concentration in the blood should remain

within a certain range that gives the desired therapeutic effects, i.e. the so-called

therapeutic range. Without a mechanism that controls the rate of release of the

drug substance, the drug concentration first rises and then decreases after

reaching its highest level, as shown in Figure 1.2. In order to better control the

drug release rate so that the drug concentration remains within the therapeutic

range for a longer period of time, formulations that give sustained or extended

drug release are developed, e.g. as shown in Figure 1.2.

Figure 1.2 Schematic of different drug release profiles

Dru

g c

oncentr

ation

Time

Toxic level

Conventional release

Sustained release

Minimum effectlevel

Therapeuticrange

1.2. The Wurster coating process 3

There are different ways of coating pellets, and among them the Wurster coating

process (Wurster, 1959) has proven to be highly efficient (Teunou & Poncelet,

2002). It is well known that the coating process that is used to produce the

controlled release film affects the release of the drug substances and therefore

the performance of the medicine. It is thus of great interest to study and

understand the Wurster coating process in greater detail.

1.2. The Wurster coating process

Figure 1.3 Schematic of the Wurster process and the different regions: 1) the spray

zone, 2) the Wurster tube, 3) the fountain region, 4) the downbed region, and 5) the

horizontal transport region.

The Wurster coating process is a bottom-spray coating process that takes place

in a fluidized bed. At the bottom of the fluidized bed, a distributor plate with a

2

3

4

5

1

Atomization

air flow

Liquid

suspensionFluidization air flow

1. Introduction 4

specific pattern of orifices determines the distribution of the hot fluidization

airflow. This airflow is partly responsible for the motion of pellets. Also, there

are one or more two-fluid spray nozzles at the bottom of the fluidized bed. These

spray nozzles supply a liquid solution as well as an atomization airflow that

breaks the solution into small droplets. The spray nozzles thus form a spray

region at the bottom of the fluidized bed. As the pellets travel through this region,

they receive a coating solution in the form of droplets, and, as they move around

in the fluidized bed, the liquid droplets are dried by the hot airflow to form a film

layer on each pellet.

Based on particle behavior, the fluidized bed can be divided into different regions,

as sketched in Figure 1.3 (Christensen & Bertelsen, 1997; Karlsson et al., 2011).

During coating pellets receive the coating solution in the spray zone, and then

dry as they return to the spray zone after having traveled through the Wurster

tube, the downbed, and the horizontal transport regions. This sequence of

coating and drying as the pellets travel through the different regions of the

fluidized bed is referred to as a particle coating cycle (Li, Rasmuson, et al., 2015).

The time it takes for a pellet to complete a cycle is defined as the cycle time.

Similarly, the time that a pellet spends in a specific region is defined as the

residence time in that region. Not only may a pellet behave differently from

other pellets, but a pellet in one cycle can also behave differently from that same

pellet in a different cycle. It is thus clear that pellets have both a cycle time

distribution (CTD) and a residence time distribution (RTD) in different regions.

The CTD and the RTD of particles in the spray zone together determine the

amount of coating deposited on the pellets. Since the performance of the

medicine to a great extent depends on the thickness of the coating film, it is clear

that the CTD and the RTD of particles in the spray zone are critical factors in

determining the coating film thickness and its variability (Mann & Crosby, 1975;

1.3. Objectives 5

Cheng & Turton, 2000; Shelukar et al., 2000; Turton, 2008). Furthermore, the

drying rate, which is affected by the RTD of particles in different regions, is also

expected to affect the quality of the coating film (Karlsson et al., 2009) and,

hence, the quality of the final product.

1.3. Objectives

Due to insufficient understanding of the underpinning mechanisms that affect

the coating quality of individual pellets in a larger population, the aim of this

study is to develop a mathematical model that is capable of predicting the

fluidized bed coating process. This was carried out in several steps, as explained

below.

The first step (Paper I (Li, Rasmuson, et al., 2015)) was to gain a basic

understanding of particle motion in the fluidized bed. In this step, the positron

emission particle tracking (PEPT) technique was employed. A series of PEPT

experiments were performed to investigate the movement of pellets and to

determine the CTD and the RTD of particles in different regions. Several

different geometrical configurations and operating parameters were considered.

In particular, different size pellets were employed to identify the underpinning

mechanisms by which the coating thickness changes.

In the second step (Paper II (Li, Remmelgas, et al., 2015)), the characteristics of

particle movement obtained, such as particle velocity, the CTD and the RTD of

particles in different regions, were selected to evaluate the model for pellet

motion based on the discrete element method, computational fluid dynamics

(DEM-CFD) simulations. In an effort (Paper III (Li et al., 2016)) to refine the

sub-models in the DEM, the effect of the drag model, which determines the

1. Introduction 6

interphase momentum transfer between the particles and the airflow, was

investigated.

In a third step (Paper II), the detailed pellet motion in the spray zone was

studied. A simple model for predicting the growth rate of pellets was developed

based on the data available from the DEM-CFD simulations and the PEPT

experiments. This predictive model was then evaluated by comparing coating

experiments carried out under similar conditions (Paper IV).

In a final step, models for the wetting and drying of pellets were developed

(Paper V). The calculated temperature and the amount of liquid in air can be

compared to the measurement at selected locations. Then it can be possible to

obtain the temperature and the amount of liquid for individual particles via

DEM-CFD simulations. In addition, varying drying rates can be identified in

different regions for an improved understanding of how the drying rate of

particles in different regions affects the coating quality.

1.4. Outline of the thesis

The thesis is organized as follows. In the next chapter, the set-up of the systems

used in this study is given to provide a basic understanding of the systems under

investigation. Chapter 3 describes the modeling methods for both the movement

and drying of pellets. In this chapter, assumptions and simplifications are

discussed and the governing equations for the modeled phenomena are given in

detail. Then the equipment, the materials, the operating conditions and the

method of analysis used in the PEPT measurements and coating experiments are

presented in Chapter 4. In Chapter 5, the main findings from the experimental

and modeling work are discussed. Last, conclusions are drawn and an outlook is

provided.

2. Set-up of the system

Prior to describing modeling methods and experiments, it is of interest to have a

basic understanding of the systems that are used in this study.

2.1. STREA-1™

For the study of particle motion, the fluidized bed was based on the STREA-1™

laboratory fluidized bed from Aeromatic-Fielder. This experimental set-up was

used for both the PEPT experiments and the DEM-CFD simulations of particle

motion.

As shown in Figure 2.1 and Table 2.1 the total height of this fluidized bed was

380 mm, and the top and bottom diameters were 250 mm and 114 mm

respectively. It was equipped with a bowl-shaped distributor with a diameter of

52 mm at the bottom of the bed. The center of this bowl-shaped distributor, i.e.

the annular region between Db and Di in Figure 2.1, was fully open and covered

the entire base. Outside this central region of the distributor, there was an

annulus region, on which a number of orifices were distributed. These orifices in

total took approximately 4% of the area of the annulus region of the distributor.

The atomization nozzle was located at the center towards the bottom of the

fluidized bed. In this study, the nozzle was only a circular orifice with a diameter

of 5 mm. No liquid solution was introduced. The diameter of the nozzle was much

larger than a real atomization nozzle in order to reduce the numerical difficulties

in simulating supersonic flow. In addition, a wire mesh screen was put over the

distributor to prevent pellets from falling through the distributor.

2. Set-up of the system 8

Figure 2.1 The dimensions of the Wurster bed, a) side view showing the different

regions: (1) the Wurster tube, (2) the fountain region, (3) the downbed region, and (4) the

horizontal transport region; b) top view.

Table 2.1 The dimensions of the STREA-1 and LAB CC fluidized bed (in mm).

Variable STREA-1 LAB CC

Diameter of expansion chamber, De 250 250

External diameter of the Wurster tube, Dt 50 50

Upper diameter of bowl distributor plate, Dc 114 100

External diameter of the base of bowl distributor plate, Db 52 -

Internal diameter of the base of bowl distributor plate, Di 12 16

Diameter of nozzle, Dn 5 -

Height of expansion chamber, He 160 730

Height of truncated cone, Hc 220 330

Height of the Wurster tube, i.e., tube length, Ht 100, 150, 200 60

Height of the connection between bowl distributor and

truncated cone, Ho 20 -

Height of bowl distributor, Hd 17 -

Height of partition gap, Hp 10, 15, 20 20

2.2. LAB CC 9

2.2. LAB CC

For the coating and drying of pellets, the LAB CC system was employed. This

system is commonly used for laboratory-scale coating experiments in

pharmaceutical development. In this research, it was employed to perform

coating and drying experiments in order to evaluate the predictive model for the

growth of pellets and the drying model under development. In the experiments,

an actual two-fluid spray nozzle was used.

This system consisted of a fluidized bed, an air supply system, a heater, sensors

and user interfaces. The fluidized bed was 1060 mm high, with bottom and top

diameters of 100 and 250 mm, respectively. Other dimensions are given in Table

2.1. It should be noted that a flat distributor was employed for the LAB CC

fluidized bed (rather than the bowl-shaped distributor that was used for the

STREA-1 fluidized bed).

3. Modeling methods

3.1. Overview of multiphase models

In modeling pellet motion in a fluidized bed, two basic approaches are available.

These two approaches, usually referred to as Eulerian-Eulerian (EE) and

Eulerian-Lagrangian (EL), are commonly used to model multiphase systems of

gas and solid particles as in this study. They are briefly described in this section.

3.1.1. Eulerian-Eulerian approach

In the EE approach, both the gas phase and particle phase are treated as

continuous, interpenetrating fluids. Since both phases are modeled as fluids, this

approach is often also referred to as the two-fluid model (TFM). The governing

equations are similar to those for single-phase flow, but a key quantity, the

volume fraction for each phase, is introduced so that the relative amount of each

phase can be determined at each location. Since only one set of conservation

equations is employed for each phase, this approach is feasible regardless of the

scale of systems. The EE approach is therefore frequently used for large-scale

systems involving many particles. Inherently, this approach provides an

averaged description of a multiphase system and requires additional relations

for coupling.

3.1.2. Eulerian-Lagrangian approach

Unlike the EE approach, particles are in the EL approach tracked as an

assembly of particles or as individual particles. In the former case, a number of

particles that are followed simultaneously are treated as parcels. In the latter

3. Modeling methods 12

case, which is the so-called discrete element method, the motion of every particle

is considered.

In pharmaceutical development, laboratory-scale systems for pellet coating often

contain a few gram to a few hundred gram almost spherical particles with a

diameter that ranges from 200 to 1000 µm. This corresponds to tens of thousands

to a few million particles, which can be simulated at a reasonable computational

cost. Pilot-scale systems contain more particles and are more challenging to

simulate. However, there is a steady increase in the computational power and

simulations of 25 million particles have been reported recently by Jajcevic et al.

(2013).

Thus, the DEM offers a unique opportunity to study pellet coating at the level of

a single pellet. With the DEM, it is feasible and reasonably straightforward to

account for different forces, including particle-particle and particle-wall collisions.

In addition, it is possible in the DEM to implement a particle size distribution

(PSD) and to study the variability in the pellet size that is usually encountered

in practice, which is difficult to account for in the EE approach. These factors

therefore motivate developing a DEM based model for pellet coating in a

laboratory-scale fluidized bed and, in the future, to extend it to pilot-scale

systems and eventually to industrial-scale systems.

3.2. Movement of pellets

In this section, the movement of pellets is described using the DEM. First, the

mathematical model for the DEM based on a soft-sphere model is presented.

Second, the governing equations for the gas flow are given. Last, the coupling

between the particle phase and the gas phase is discussed.

3.2. Movement of pellets 13

3.2.1. Discrete element method

Basically, the DEM is based on Newton’s second law of motion. The motion of

every pellet is then written via a force balance:

𝑚𝑝,𝑖

𝑑𝒗𝒑,𝒊

𝑑𝑡= 𝛽

𝑉𝑝,𝑖

휀𝑠(𝒖𝒈 − 𝒗𝒑,𝒊) + 𝑚𝑝,𝑖𝒈 − 𝑉𝑝,𝑖𝛻𝑃 + 𝑭𝒄,𝒊 (3.1)

where the terms on the RHS represent, in order, interphase momentum transfer,

gravity, the force due to the pressure gradient, and the contact force due to

particle-particle and particle-wall collisions. In this equation, 𝑚𝑝,𝑖 is the mass of

the 𝑖𝑡ℎ pellet, 𝒗𝑝,𝑖 is the velocity of the 𝑖𝑡ℎ pellet, 𝒖𝑔 is the velocity of the air, 𝒈 is

the gravitational acceleration, 𝑉𝑝,𝑖 is the volume of the 𝑖𝑡ℎ pellet, 𝛻𝑃 is the

gradient of the air pressure, 휀𝑠 is the particle/solid volume fraction, and 𝛽 is the

interphase momentum transfer coefficient.

The angular momentum of the pellet is calculated by

𝐼𝑝,𝑖𝑑𝝎𝒑,𝒊

𝑑𝑡=∑𝑻𝑝,𝑖 (3.2)

where 𝐼𝑝,𝑖 is the moment of inertia, 𝝎𝑝,𝑖 is the rotational velocity, and 𝑻𝒑,𝒊 is the

total torque acting on the particle.

For the contact force due to particle-particle and particle-wall collisions, it is

well-known that there are two models available: the hard sphere and soft sphere

models. In the hard sphere model, the interaction forces are assumed to be

impulsive and all other forces are negligible during collisions (Crowe et al., 2012).

The hard sphere model is easy to use but limited to only binary collisions. In the

soft sphere model, the deformation of particles in contact is modeled in a

straightforward fashion using the Hertz-Mindlin theory (Hertz, 1882); the local

linear deformation is related to the normal and tangential forces respectively.


The soft sphere model implements the physics of collisions more directly but at a

higher computational cost. In this study, since pellets can remain in contact for a

long time in the dense region of the fluidized bed, the soft sphere model has been

used, as described below.

i j

Spring

DashpotSlider

Figure 3.1 A schematic of the soft sphere model

When particles collide with each other, their deformation is modeled via an

overlap between them and the energy loss during the collision is accounted for

through a so-called spring-dashpot system, as illustrated in Figure 3.1. This

system is characterized using the spring stiffness, 𝑘, the damping coefficient, 𝜂,

and the friction coefficient, 𝜇𝑝. The latter quantity (𝜇𝑝) is empirical and can be

measured by e.g. using a Jenike shear cell (Jenike, 1961; Darelius et al., 2007).

The former two quantities can be calculated according to the Hertz-Mindlin

theory (Hertz, 1882; Dintwa et al., 2007), as explained below.

The normal and tangential contact forces acting on the particle, 𝑭𝒄𝒏,𝒊𝒋 and 𝑭𝒄𝒕,𝒊𝒋,

are given by (Crowe et al., 2012)

𝑭𝒄𝒏,𝒊𝒋 = −(𝑘𝑛𝛿𝑛3/2𝒏𝑖𝑗 − 𝜂𝑛𝑗𝒗𝑖𝑗 ∙ 𝒏𝑖𝑗)𝒏𝑖𝑗 (3.3)

𝑭𝑐𝑡,𝑖𝑗 = {−𝑘𝑡𝜹𝑡 − 𝜂𝑡𝑗𝒗𝑐𝑡 𝑖𝑓 |𝑭𝑐𝑡,𝑖𝑗| ≤ 𝜇𝑝|𝑭𝑐𝑛,𝑖𝑗|

−𝜇𝑝|𝑭𝑐𝑛,𝑖𝑗|𝒕𝑖𝑗 𝑖𝑓 |𝑭𝑐𝑡,𝑖𝑗| > 𝜇𝑝|𝑭𝑐𝑛,𝑖𝑗| (3.4)


where 𝛿 represents the displacement of the particle caused by the normal or

tangential force, 𝒏𝑖𝑗 and 𝒕𝑖𝑗 are the unit vectors normal and tangential to the

contact plane, respectively, 𝒗𝑖𝑗 is the relative velocity between particle 𝑖 and

particle 𝑗, and 𝒗𝑐𝑡 is the slip velocity of the contact point. The suffixes 𝑛 and 𝑡

denote the components in the normal and tangential directions, while the

suffixes 𝑖 and 𝑗 denote particle 𝑖 and particle 𝑗, respectively.

The normal and tangential stiffnesses, 𝑘𝑛 and 𝑘𝑡, are expressed by

𝑘𝑛 =4

3(1 − 𝜎𝑖

2

𝐸𝑖+1 − 𝜎𝑗

2

𝐸𝑗)

−1

(𝑟𝑖 + 𝑟𝑗

𝑟𝑖𝑟𝑗)

−1 2⁄

(3.5)

𝑘𝑡 = 8(1 − 𝜎𝑖

2

𝐻𝑖+1 − 𝜎𝑗

2

𝐻𝑗)

−1

(𝑟𝑖 + 𝑟𝑗

𝑟𝑖𝑟𝑗)

−1 2⁄

𝛿𝑛1/2 (3.6)

where 𝐸 is the Young’s modulus, 𝜎 is the Poisson’s ratio, 𝐻 =𝐸

2(1+𝜎) is the shear

modulus and 𝑟 is the radius of the particle.

In Eq. (3.3) and Eq. (3.4), the damping coefficient 𝜂, which represents the viscous

dissipation of kinetic energy in the normal and tangential directions, can be

obtained according to Tsuji et al. (1992); (1993),

𝜂𝑛 = 2𝛼√𝑚𝑝∗𝑘𝑛𝛿𝑛

1/4 (3.7)

𝜂𝑡 = 2𝛼√𝑚𝑝∗𝑘𝑡 (3.8)

where 𝛼 denotes a constant related to the coefficient of restitution, which is given

in the work of Tsuji et al. (1992), and 𝑚𝑝∗ represents the effective particle mass

and is calculated by

𝑚𝑝∗ =

𝑚𝑝,𝑖𝑚𝑝,𝑗

𝑚𝑝,𝑖 +𝑚𝑝,𝑗 (3.9)

A more detailed description of the model can be found in the literature (Cundall

& Strack, 1979; Tsuji et al., 1993; Deen et al., 2007; Crowe et al., 2012).


3.2.2. Gas flow

As mentioned earlier, there is an atomization airflow with a high velocity at the

bottom of the fluidized bed. From the atomization airflow rate and the diameter

of the nozzle, the Reynolds number is estimated to be approximately 104. For

single-phase airflow with such high Reynolds number, turbulence may therefore

be expected. Nevertheless, since the Stokes number of the large particles in this

study is quite large, St>>1, it may be assumed that turbulence near the nozzle

will not have an appreciable effect on particle trajectories. Additionally, for dense

gas-solid flows, such as the one included in this study, the particle stress is

expected to be much greater than the stress due to turbulence (Hrenya &

Sinclair, 1997). Turbulence is therefore neglected.

Therefore, the continuity and momentum equations for the airflow can be

written:

𝜕

𝜕𝑡(휀𝑔𝜌𝑔) + 𝛻 ∙ (휀𝑔𝜌𝑔𝒖𝑔) = 0 (3.10)

𝜕

𝜕𝑡(휀𝑔𝜌𝑔𝒖𝑔) + 𝛻 ∙ (휀𝑔𝜌𝑔𝒖𝑔𝒖𝑔)

= −휀𝑔𝛻𝑃 − 𝛻 ∙ (휀𝑔𝝉𝑔) + 휀𝑔𝜌𝑔𝒈 −∑𝛽(𝒖𝑔 − 𝒗𝑝,𝑖)

𝑛

𝑖=1

(3.11)

where 휀𝑔 = 1 − 휀𝑠 is the gas volume fraction, 𝜌𝑔 is the density of the air, 𝜇𝑔 is the

dynamic viscosity of the air, 𝑑𝑝 is the particle diameter and 𝝉𝒈 is the viscous

stress tensor for incompressible flow,

𝝉 𝑔 = 𝜇𝑔 (𝛻𝒖𝑔 + (𝛻𝒖𝑔)𝑇) (3.12)

3.2.3. Drag model

The momentum transfer between the particles and the air is given by the

interphase momentum transfer term as given in Eqs. (3.1) and (3.11), which is


determined by the interphase momentum transfer coefficient, 𝛽. This interphase

momentum transfer coefficient can be specified using different drag models. This

study mainly relies on the Gidaspow drag model (Gidaspow, 1994), but also

explores four other models for the possibility of refinement: the Min, HKL,

Beetstra and Tang drag models. The latter three models are based on

simulations of flow through assemblies of particles using the Lattice-Boltzmann

method (LBM) or the immersed boundary method (IBM) while the Min drag

model is merely the minimum of the Ergun and Wen-Yu drag models. These drag

models give different interphase momentum transfer coefficients, as shown in

Figure 3.2, for varying relative velocities and volume fractions. Details about the

Gidaspow drag model are given below, while details about the other four drag

models can be found in Paper III.

The Gidaspow drag model is based on the Ergun equation (Ergun, 1952) for the

dense regime and the Wen and Yu correlation (Wen & Yu, 1966) for the dilute

regime,

𝛽𝐺𝑖𝑑𝑎𝑠𝑝𝑜𝑤 =

{

𝛽𝐸𝑟𝑔𝑢𝑛 = (150

휀𝑠2

휀𝑔+ 1.75휀𝑠𝑅𝑒𝑝)

𝜇𝑔

𝑑𝑝2 휀𝑠 > 0.2

𝛽𝑊𝑒𝑛−𝑌𝑢 =3

4𝑅𝑒𝑝𝐶𝐷

𝜇𝑔휀𝑠

𝑑𝑝2 휀𝑔

−2.65 휀𝑠 ≤ 0.2

(3.13)

where 𝐶𝐷 is the drag coefficient, which is written as

𝐶𝐷 = {24(

1 + 0.15𝑅𝑒𝑝0.687

𝑅𝑒𝑝) 𝑅𝑒𝑝 < 1000

0.44 𝑅𝑒𝑝 ≥ 1000

(3.14)

and 𝑅𝑒𝑝 is the particle Reynolds number,

𝑅𝑒𝑝 = 휀𝑔𝜌𝑔|𝒖𝑔 − 𝒗𝑝,𝑖|𝑑𝑝

𝜇𝑔 (3.15)


This drag model was found to be the best for modeling spouted beds (Du et al.,

2006) and has been widely used in engineering practice (Deen et al., 2007).

Figure 3.2 Comparison of drag models, the relative velocity between gas flow and

particles is (a) 5 m/s and (b) 1 m/s, respectively.

0

0.2

0.4

0.6

0.8

1

0 0.05 0.1 0.15 0.2

β(1

03 k

g/(

m3.s

))

Particle volume fraction (-)

β,Gidaspow

Min(β,Ergun, β,Wen-Yu)

β, HKL,extension

β, Beetstra

β, Tang

a

0

0.5

1

1.5

2

0.2 0.3 0.4 0.5 0.6

β(1

03 k

g/(

m3.s

))

Particle volume fraction (-)

β,Gidaspow

Min(β,Ergun, β,Wen-Yu)

β, HKL,extension

β, Beetstra

β, Tang

b

3.3. Numerical methods 19

3.3. Numerical methods

To solve the governing equations presented above, a fully-coupled multiphase

solver, MultiFlow (http://www.multiflow.org), developed by van Wachem’s

research group, was employed. In MultiFlow, the continuous phase is predicted

using the finite volume method (FVM) via the Rhie and Chow (1983)

interpolation for pressure and velocity coupling. For spatial discretization, the

second-order central difference scheme is used for non-convective terms while

the second-order upwind scheme is used for convective terms. The time

discretization is based on the second-order backward scheme (Jorn Bruchmüller,

2011).

It is worth mentioning that a multigrid technique was employed in this research

to couple the discrete and continuous phases as described by e.g. J. Bruchmüller

et al. (2010); (2011). With this technique, the continuous phase is solved on a

hexahedral mesh while particles are tracked on a so-called particle mesh, which

is a Cartesian mesh with a mesh size that is larger than the particle. All the

coupling terms, including the volume fraction and the drag force, are determined

on the length-scale of this particle mesh. In particular, if the local fluid cell is

smaller than a particle, the coupling between that fluid cell and the particle

occupying that fluid cell is length-scale weighted, and therefore remain physical.

This way of addressing the very large volume fraction present in dense particle

flow has been described and tested by Mallouppas and van Wachem (2013), and

has been further worked out in the recent work of Capecelatro and Desjardins

(2013).

In terms of boundary conditions, a velocity inlet normal to the boundary was

applied at the distributor and at the spray nozzle. At the outlet of the fluidized

bed, a pressure outlet was employed. In addition, the walls of the fluidized bed

http://www.multiflow.org/


and the Wurster tube, which are treated as adiabatic, were set to be no-slip for

the continuous phase.

3.4. Drying of pellets

As mentioned earlier, for a given polymer film, the coating quality not only

depends on the film thickness but can also depends on the drying rate of the

liquid film. In this study, it is assumed that as a pellet passes through the spray

zone and receives the coating solution, an evenly spread liquid film is formed

quickly around the surface of the pellet. As the pellet leaves the spray zone, it is

dried by the hot airflow as described as follows.

Mass conservation of liquid in pellets:

𝜕𝑚𝑝,𝑤

𝜕𝑡= 𝐴𝑝𝑁𝑤,𝑝 (3.16)

where 𝐴p is the surface area of the pellet, 𝑁𝑤,𝑝 is the mass flux of liquid through

the surface of the pellet, 𝑚𝑝,𝑤 is the mass of the pellet. Here the total mass of the

pellet is the sum of the mass of the solid core, 𝑚𝑝, the mass of the solvent, 𝑚𝑤,

and the mass of the polymer substance, 𝑚𝑠𝑢𝑏:

𝑚𝑝,𝑤 = 𝑚𝑝 +𝑚𝑤 +𝑚𝑠𝑢𝑏 (3.17)

Eq. (3.16) can then be written as

𝜕𝑚𝑤

𝜕𝑡= 𝐴𝑝𝑁𝑤,𝑝 (3.18)

For simplification, the following assumptions are made in a first model attempt:

i. In the first drying stage, the pellet surface is saturated with liquid;

3.4. Drying of pellets 21

ii. In the intermediate drying stage, the drying rate falls mainly due to

the decreased vapor pressure at equilibrium. In reality, the drying

rate can also decrease due to an increase in the mass transfer

resistance, but this effect is not taken into account.

iii. The change in the film thickness does not affect the hydrodynamics

of pellet motion.

The mass flux of liquid to the surface of the pellet during drying can be written

as

𝑁𝑤,𝑝 =𝑘𝑝𝑔𝑀𝑤

𝑅(𝑃𝑤,𝑝

𝑇𝑝−𝑃𝑤,𝑔

𝑇𝑔) (3.19)

In Eq. (3.19), 𝑘𝑝𝑔 is the mass transfer coefficient, 𝑀𝑤 is the molar weight of the

solvent, 𝑅 is the gas constant, 𝑇𝑝 and 𝑇𝑔 are the temperature of the pellet and the

temperature of the air, respectively, 𝑃𝑤,𝑔 is the partial pressure of the solvent in

air and 𝑃𝑤,𝑝 is the partial pressure of the solvent over the surface of the pellet.

In the first stage, 𝑃𝑤,𝑝 can be calculated according to the Antoine equation for the

saturated vapor of the solvent, 𝑃𝑤,𝑠𝑎𝑡,

𝑃𝑤,𝑝 = 𝑃𝑤,𝑠𝑎𝑡 =101325

76010

(𝐴−𝐵

𝑇𝑝−273.15+𝐶) (3.20)

where 𝐴, 𝐵 and 𝐶 are constants for the specified solvent, such as water in this

study. In Eq. (3.20), the units for pressure and temperature are Pascal and

Kelvin, respectively.

In the second stage, the decrease in driving force caused by the lower vapor

pressure at equilibrium with the polymer film is accounted for. In this case, the

partial pressure of the solvent over the surface of the pellet can be related to the

saturated vapor of the solvent,


𝑃𝑤,𝑝 = 𝑓(𝑋) ∗ 𝑃𝑤,𝑠𝑎𝑡 (3.21)

where 𝑓(𝑋) is a function of the mass fraction of the solvent in the pellet, 𝑌𝑤,𝑝,

𝑓(𝑋) = 100(1 − 𝑒𝑥𝑝 (−𝑘 𝑌𝑤,𝑝𝑁)) (3.22)

where 𝑘 and 𝑁 are constants that can be determined by fitting the experimental

data for the specified material. In Eq. (3.22), 𝑌𝑤,𝑝 can also be related to the mass

fraction of the solvent in the coating film, for example, as shown in Figure 3.3.

Figure 3.3 The relation between relative humidity and moisture content, e.g. for

HPMC.

In Eq. (3.19), the correlations proposed by Gunn (1978) are used to calculate the

mass transfer coefficient:

𝑘𝑝𝑔 =𝐷𝐴𝐵𝑆ℎ𝑝

𝑑𝑝 (3.23)

𝑆ℎ𝑝 = (7 − 10휀𝑔 + 5휀𝑔2)(1 + 0.7𝑅𝑒𝑝

0.2𝑆𝑐1/3)

+ (1.33 − 2.4휀𝑔 + 1.2휀𝑔2)𝑅𝑒𝑝

0.7𝑆𝑐1/3 (3.24)

𝑆𝑐 =𝜇𝑔

𝜌𝑔𝐷𝐴𝐵 (3.25)

0

20

40

60

80

100

0 5 10 15 20 25 30 35

Rela

tive h

um

idity [

%]

Moisture content [%]

exp.

fit

3.4. Drying of pellets 23

where 𝑆ℎ𝑝 is the Sherwood number of the pellet, 𝑆𝑐 is the Schmidt number of the

air, 𝐷𝐴𝐵 is the diffusivity of the solvent in air.

Then the mass fraction of the solvent, 𝑌𝑤,𝑝, can be related to the film thickness,

𝑙𝑓𝑖𝑙𝑚:

𝑌𝑤,𝑝 =𝑚𝑤

𝑚𝑝=𝐴𝑝𝑙𝑓𝑖𝑙𝑚𝜙1𝜌𝑤

𝑚𝑝 (3.26)

where 𝑚𝑤 is the mass of the solvent, 𝜙1 is the volume fraction of the solvent and

𝜌𝑤 is the density of the solvent.

Energy equation over pellets:

𝑚𝑝𝑐𝑝,𝑝𝜕𝑇𝑝

𝜕𝑡= −ℎ𝑝𝑔𝐴𝑝(𝑇𝑝 − 𝑇�̃�) − 𝐻𝑒𝑣𝑎𝑝𝑆𝑤,𝑝 (3.27)

where the first and second terms on the RHS represent the convective heat

transfer and the heat transfer due to evaporation of the solvent. In Eq. (3.27), 𝑐𝑝,𝑝

is the specific heat capacity of the pellet, ℎ𝑝𝑔 is the convective heat transfer

coefficient estimated from the Nusselt number, which has a form similar to Eq.

(3.23), 𝑇�̃� is the interpolated fluid temperature as seen by the pellet, 𝐻𝑒𝑣𝑝 is the

latent heat of evaporation and 𝑆𝑤,𝑝 =𝜕𝑚𝑤

𝜕𝑡 is the evaporation rate.

Mass conservation of solvent for the fluid phase:

𝜕휀𝑔𝜌𝑔𝑌𝑤,𝑔

𝜕𝑡+ 𝛻 ∙ (휀𝑔𝜌𝑔𝒖𝒈𝑌𝑤,𝑔) = 𝛻 ∙ (휀𝑔𝐷𝐴𝐵𝛻𝑌𝑤,𝑔) − 𝑆𝑤,𝑔 (3.28)

where 𝑌𝑤,𝑔 is the mass fraction of the solvent in air. Here, 𝑆𝑤,𝑔 is the solvent

exchange rate between the air and the particles:

𝑆𝑤,𝑔 = −∑𝑆𝑤,𝑝 (3.29)


Energy equation for the fluid phase:

𝜕휀𝑔𝜌𝑔𝑐𝑝,𝑔𝑇𝑔

𝜕𝑡+ 𝛻 ∙ (휀𝑔𝜌𝑔𝒖𝒈𝑐𝑝,𝑔𝑇𝑔) = 𝛻 ∙ (휀𝑔𝜆𝑔𝛻𝑇𝑔) + 𝑄𝑝𝑔 + 𝐻𝑒𝑣𝑎𝑝𝑆𝑤,𝑔 (3.30)

where 𝑐𝑝,𝑔 is the heat capacity of the air and 𝜆𝑔 is the thermal conductivity of the

air.

In Eq. (3.30), 𝑄𝑝𝑔 is the heat exchange rate between the air and the particles:

𝑄𝑔 =∑6휀𝑔ℎ𝑝𝑔(𝑇𝑝 − 𝑇𝑔)

𝑑𝑝 (3.31)

4. Experimental

4.1. PEPT

4.1.1. The PEPT measurement system

Figure 4.1 Schematic of the PEPT measurement system

As illustrated in Figure 4.1, the PEPT measurement system consisted of a

fluidized bed (containing particles) in the middle, two position-sensitive detectors

on both sides of the fluidized bed and a tracer particle. The tracer particle was

marked with a radioactive isotope that releases γ-rays. As the tracer particle

moves around in the equipment, the γ-rays are captured by the detectors. Then it

is possible to define a line of response (LOR) along which the tracer particle lies

by following the pair of γ-rays detected simultaneously. After a sufficient number

of pairs of γ-rays have been obtained, the location of the tracer particle in three

dimensions can be found through geometric triangulation. More details about the

4. Experimental 26

technique and the algorithm have been presented elsewhere by Parker et al.

(1993); (1997; 2002).

4.1.2. Materials

In this study, microcrystalline cellulose (MCC) pellets, which are commonly-used

solid cores in the pharmaceutical industry, were employed. These MCC pellets,

manufactured upon request by Umang Pharmatech Pvt. Ltd, are relatively large

in order to limit the number of particles in the DEM-CFD simulations. The

pellets were classified into two main groups according to their size distribution

and the corresponding volume mean diameter (VMD): small (approximately 1.8

mm) and large (approximately 2.6 mm).

4.1.3. The tracer particle

Prior to the experiments, using a scanning electron microscope (SEM), a number

of pellets with a diameter close to the VMD were selected from the respective

small and large pellets for later labeling use. In each run, only one tracer particle

was incorporated. This tracer particle was put in radioactive water for about half

an hour so as to gain free 18F- via absorption and adsorption. Since the

radioactivity lasts for about 6 hours, the same tracer particle could be used for

several runs. Since the activity of the tracer particle decreases as a function of

time, it was necessary to move the detectors closer to the fluidized bed in later

runs.

4.1.4. Operating parameters

In the PEPT experiments, a suitable airflow rate was selected based on several

preliminary tests. The objective of these tests was to find a flow rate that gave a

desired representative of the pellet motion in a Wurster fluidized bed, as

4.1. PEPT 27

described in the introduction. For a typical run, 200 g monosized small pellets

were employed, while for mixtures, 25%, 50% and 75%, the 200 g particles were

replaced with the large ones. In the preliminary tests, it was also found that a

run time of 1.5 hours was sufficient to obtain a compromise between the quality

of the data and the run time. A summary of the configurations and operating

parameters can be found in Table 4.1 for each run. Run #2 was selected as the

base case because its operating parameters were set to the middle level.

Table 4.1 The configuration and operating parameters for each PEPT run.

Run

#

Parti-

tion

gap

(mm)

Batch

size

(g)

Tube

length

(mm)

Mass of

1749 µm

particles

(g)

Mass of

2665 µm

particles

(g)

VMD of

tracer

particle

(µm)

Fluidization

airflow rate

(m3/h)

Atomization

airflow rate

(m3/h)

Run

time

(h)

1 10 200 150 200 - 1749 73.3 3.50 3

2 15 200 150 200 - 1749 73.3 3.50 1.5

3 20 200 150 200 - 1749 73.3 3.50 1.5

4 15 400 150 400 - 1749 73.3 3.50 1.5

5 15 600 150 600 - 1749 73.3 3.50 1.5

6 15 200 100 200 - 1749 73.3 3.50 1.5

7 15 200 200 200 - 1749 73.3 3.50 1.5

8 15 200 150 - 200 2665 80.3 4.32 1.5

9 15 200 150 150 50 1749 80.3 4.32 1.5

10 15 200 150 150 50 2665 80.3 4.32 1.5

11 15 200 150 100 100 1749 80.3 4.32 1.5

12 15 200 150 100 100 2665 80.3 4.32 1.5

13 15 200 150 50 150 1749 80.3 4.32 1.5

14 15 200 150 50 150 2665 80.3 4.32 1.5

15 15 200 150 200 - 1749 65.8 3.50 1

16 15 200 150 200 - 1749 80.3 3.50 1

17 15 200 150 200 - 1749 73.3 2.49 1

18 15 200 150 200 - 1749 73.3 4.54 1

4.1.5. Post-processing

In the PEPT experiments, the location of the tracer particle was obtained using

the program, TRACK, which was developed by Parker et al. (1993). In the

4. Experimental 28

algorithm used by the program, the events recorded by the PEPT cameras are

divided into subsets of sequential events by specifying the number of events per

slice, #events/slice. The location that minimizes the distance to the reconstructed

paths is then calculated for each subset. As a next step, the events farthest from

the calculated location are regarded as corrupted and are discarded. This

procedure of discarding corrupted events is repeated until a certain fraction of

the sequential events for each subset, 𝑓𝑜𝑝𝑡, remain, at which point the location of

the tracer particle is found. Then it is possible to calculate the average distance

of all the LORs used to determine the location of the tracer particle from the

calculated location. This average distance, i.e. error, then provides a measure of

precision or reliability of that calculated location (Seville, 2010).

Figure 4.2 An example showing the axial (y) coordinates of the tracer particle

positioned on top of the nozzle with varying errors.

The optimal values of these parameters depend on the particular geometries and

flow conditions (Chiti, 2008). For the data obtained here, an example of the

0

100

200

300

400

500

600

80 85 90 95 100 105 110

Y c

oo

rdin

ate

(m

m)

Run time (s)

error: 10

error: 5

error: 2

4.1. PEPT 29

removal of outliers with a varying error can be seen in Figure 4.2. Figure 4.2

shows that an error of 2 removes most of the outliers, but that it also greatly

reduces the number of data points available. Thus, there is a trade-off between

the accuracy or error and the number of data points obtained. High values for

the number of events per slice, 𝑓𝑜𝑝𝑡 and the error give a higher time resolution

(i.e. more data points) but a lower reliability. On the other hand, low values of

the number of events per slice, 𝑓𝑜𝑝𝑡, and the error give a lower time resolution

but a higher reliability. After a number of tests, values for the number of events

per slice, 𝑓𝑜𝑝𝑡 and the error that gave a sufficient time resolution and reliability

were determined to be 500, 0.05 and 5, respectively. Using these parameter

values, the sampling interval in this study remained in the range of 50-80 ms.

In this range, the effect of these parameter values on e.g. the cycle time is

relatively minor, as shown in Figure 4.3.

Figure 4.3 The effect of the mean sampling interval on the mean cycle time (partition

gap 10 mm, batch size 200 g, particle VMD 1749 μm, tube length 150 mm, fluidization

airflow rate 73.3 m3/h, and atomization airflow rate 3.50 m3/h).

8.05

7.70 7.69 7.84

8.44

9.46

5

6

7

8

9

10

0 50 100 150 200

Mean

cycle

tim

e (

s)

Sampling interval (ms)

4. Experimental 30

4.2. Coating experiments

4.2.1. Materials

The pellets used in the coating experiments were the same as those used in the

PEPT experiments. These pellets were mixed so as to produce the same total

amount of mixtures with different fractions of large particles: 25%, 50% and 75%.

A detailed composition of each mixture and the VMD of small and large pellets in

each mixture are presented in Table 4.2.

Table 4.2 Specification of uncoated small and large particles for different mixtures.

Fraction of the large particles

in each mixture

(%)

Mass of

the small

particles

(g)

Mass of

the large

particles

(g)

Class of particles

(-)

VMD

(µm)

25 135 45 Small 1783

Large 2704

50 90 90 Small 1768

Large 2694

75 45 135 Small 1780

Large 2675

4.2.2. QicPic analysis

In order to monitor the change in size of uncoated and coated pellets, an image

analyzer, namely QicPic, was employed. With QicPic, both the size and shape of

particles can be measured. In this measurement, the diameter of a pellet is

defined as the diameter of a circle that has the same area as the projection area

of the pellet. The sphericity is specified as the ratio of the perimeter of the

equivalent circle to the real perimeter. In order to be able to collect the pellets

after the measurement, the GRADIS dispenser was employed to ensure that the

measurement was non-destructive.

4.2. Coating experiments 31

4.2.3. Coating solution

The pellets were coated with a polymer solution containing ethyl cellulose (EC)

and water-soluble hydroxypropyl cellulose (HPC). The coating solution was

prepared by dissolving EC/HPC in ethanol at room temperature followed by

stirring overnight. In this study, the EC/HPC polymer blend ratio was 50:50 w/w

and the polymer concentration was 6% w/w.

4.2.4. Determination of growth of pellets

In this study, the size range of small and large pellets had an overlap to some

extent. Since this overlap may cause difficulty in calculating the respective

growth of small and large pellets, i.e. in determining whether a pellet is small or

large, it was necessary to minimize the overlap. Therefore, screens with

apertures of 2 mm and 2.5 mm were employed to sieve the pellets larger than 2

mm from the small pellets and the pellets smaller than 2.5 mm from the large

pellets, respectively. Then mixtures of small and large pellets were prepared

according to the specifications in Table 4.2. Since the small and large pellets

have their own size distribution, the small and large pellets for each mixture

were measured separately before mixing. Afterwards, the overall PSD of each

mixture was measured.

After coating, it was found that the PSD of the mixtures depended very strongly

on the sample size. Since it was impractical to measure all particles at the same

time, each mixture was measured by splitting it into three parts and then

calculating the overall PSD using a mass weighted average. In order to evaluate

the growth of pellets after coating, a threshold particle size was selected to

distinguish the small and large particles in each mixture. The threshold particle

size for each mixture was specified as the particle size between 2000 and 2500

µm for which the weight fraction was the smallest.

4. Experimental 32

4.3. Drying experiments

4.3.1. Method

In the drying experiments, the built-in sensors in the LAB CC system read the

temperature and the humidity at the inlet and the outlet of the fluidized bed. In

addition, six sensors, namely PyroButton®, are used to log the local humidity

and the temperature. The PyroButton® has recently been utilized in

pharmaceutical industry to improve understanding and assist process operation

(Kona et al., 2013; Pandey & Bindra, 2013). In this study, these sensors were

employed to record the relative humidity (RH) and the temperature at selected

locations in the Wurster tube.

4.3.2. Coating solution

In the drying experiments, the polymer solution with a polymer concentration of

6% was produced by dissolving and stirring HPMC in water for 24 hours. For

each experiment, 20% weight increase to the core pellets was planned, which

resulted in a consumption of approximately 730 g coating solution.

4.3.3. Operating parameters

In the drying experiments, small and large microcrystalline cellulose (MCC)

pellets with diameters of 1.8 and 2.6 mm were employed. Operating parameters,

i.e. fluidization airflow rate, air temperature and spray rate, were varied and

summarized in Table 4.3. The fluidization and atomization air was dry. It is of

interest to note that due to practical reasons, the highest spray rate was limited

to approximately 18 g/min. Every run took approximately 40 minutes for coating

and 2 minutes for additional drying after the coating solution was stopped. Since

the PyroButtons® also receive an amount of coating solution, they were cleaned

between runs. In addition, pure water was used in Run #8 to provide a reference.

4.3. Drying experiments 33

Table 4.3 The operating parameters used in the drying experiments.

Run

#

Mass of

1.8 mm

pellets

(g)

Mass of

2.6 mm

pellets

(g)

Fluidization

airflow rate

(m3/h)

Atomization

airflow rate

(m3/h)

Air

temperature

(°C)

Spray

rate

(g/min)

1 - 175 60

2.5

75

18

2 175 -

3

- 175

50

4 70

5

60

60

6 90

7 175 - 75

10

8 18

5. Results and discussion

5.1. Characteristics of particle motion (Paper I)

Figure 5.1 Different types of trajectories: (a) cycle, and (b) recirculation. The bold lines

are the walls of the fluidized bed and the Wurster tube.

As described in the last section, a sequence of locations of the tracer particle was

stored in each PEPT run. By following the location of the tracer particle over

time, two types of trajectories can be defined:

a. Cycle (Figure 5.1a): the tracer particle departs from the spray zone,

accelerates in the Wurster tube, and returns to the spray zone after

passing through the fountain, downbed and horizontal transport regions.

b. Recirculation (Figure 5.1b): the tracer particle recirculates within the

Wurster tube without passing through the downbed region. This type of

trajectory is not desired because it is believed to increase the risk for

0

50

100

150

200

250

300

0 25 50 75 100 125

Ax

ial

dis

tan

ce (

mm

)

Radial distance (mm)a)

0

50

100

150

200

250

300

0 25 50 75 100 125

Ax

ial

dis

tan

ce (

mm

)

Radial distance (mm)b)

5. Results and discussion 36

agglomeration of particles and to give a broader film thickness

distribution.

It is noted that since it was not possible to define a spray region in the PEPT

experiments, a particle cycle begins when first appearing in the Wurster tube,

rather than in the spray zone.

Figure 5.2 Residence time distributions in specific regions, with mean values of all

cycles and ideal cycles provided (partition gap 15 mm, batch size 200 g, particle VMD

1749 μm, tube length 150 mm, fluidization airflow rate 73.3 m3/h, and atomization

airflow rate 3.50 m3/h), the bin size for the fountain and the downbed regions: 0.1 s, for

the Wurster tube: 0.5 s, and for the horizontal transport region: 1 s.

In a real coating process, it is desirable to minimize particle recirculations.

Nevertheless, it was observed in the PEPT experiments that the particle tends to

recirculate in the Wurster tube. When a cycle does not have any recirculation, it

is regarded as an ideal cycle, otherwise it is non-ideal.

0

50

100

150

200

250

0 5 10 15

Nu

mb

er

of

cycle

s

Residence time in specific regions (s)

The fountain region, mean: 0.27/0.26s (all/ideal cycles)

The downbed region, mean: 0.33/0.11 s (all/ideal cycles)

The Wurster tube, mean: 1.38/0.52 s (all/ideal cycles)

The horizontal transport region, mean: 8.16s/2.15 s (all/ideal cycles)

5.1. Characteristics of particle motion (Paper I) 37

Figure 5.3 An example of the axial (above) and radial (below) position of a tracer

particle during 70 s.

Figure 5.2 shows the RTDs of particles in different regions from all cycles and

ideal cycles. It can be seen that the particle on average spends the longest time,

0

50

100

150

200

250

300

230 240 250 260 270 280 290 300

Ax

ial

dis

tan

ce

(m

m)

Time (s)

Wu

rste

rtu

be

0

25

50

75

100

125

230 240 250 260 270 280 290 300

Ra

dia

l d

ista

nce

(m

m)

Time (s)

Wu

rste

r tub

e


8.16 s, in the horizontal transport region, while it spends a much shorter time,

1.38 s, 0.27 s, and 0.33 s in the Wurster tube and in the fountain and downbed

regions, respectively. Moreover, the residence time in the Wurster tube and

particularly in the horizontal transport region is much shorter for the ideal

cycles than for all cycles. In principle, the main difference between ideal and

non-ideal cycles is the particle recirculation that occurs in the Wurster tube.

However, it was found that the very long residence time in the horizontal

transport region for non-ideal cycles most likely is a result of the particle

returning to the horizontal transport region by passing out of the Wurster tube

through its lower edge. This conclusion is supported by an example of particle

trajectory during the 230-300 s as shown in Figure 5.3, more specifically at 238 s

and 263 s.

More effects, such as the effects of the partition gap, the batch size, the length of

the Wurster tube, the fluidization and atomization airflow rates, and binary

mixtures on pellet motion were also investigated in Paper I. For a given piece of

equipment, it was found that it is possible to optimize the partition gap so that it

is small enough to direct the fluidization airflow into the Wurster tube and large

enough not to limit the solids flux into the tube. It was also observed that, as the

batch size increases, the cycle time decreases and the CTD becomes narrower.

This implies that it is possible to coat a larger batch of pellets in a shorter

amount of time and while ensuring less variability in the coating film thickness.

Nevertheless, this may be limited by the spray rate since a higher spray rate

then is required for a larger batch. This, however, suggests opportunities to

establish more efficient processes by exploring higher spray rates for larger

batches.

For binary mixtures of small and large pellets, it was found that large pellets

have a longer cycle time than small pellets, which indicates that smaller pellets

5.2. Validation and development of DEM-CFD model (Paper II) 39

get coated more frequently. Nevertheless, it was not clear in previous work

(Sudsakorn & Turton, 2000; Paulo Filho et al., 2006; Cahyadi et al., 2012;

Marucci et al., 2012) whether large pellets would receive a higher total amount of

coating mass than small pellets because large pellets circulated more often or

because large pellets received more coating when they passed through the spray

zone. The result here clearly provides evidence for the latter effect.

5.2. Validation and development of DEM-CFD model

(Paper II)

5.2.1. Calibration of inlet flow

As mentioned earlier, the velocity at the inlet boundary at the distributor is

determined by fluidization airflow and the velocity at the nozzle depends on

atomization airflow. For fluidization airflow, which has a major effect on how the

particles behave in the fluidized bed (Li et al., 2016), it is necessary to specify

airflow distribution. Since it was not possible to experimentally measure airflow

distribution, several DEM-CFD simulations using various flow distributions

were performed. The resulting particle velocities at different heights were

compared with the PEPT measured values. Then the flow distribution, under

which the adequate agreement between the calculated and measured particle

velocity, was achieved is defined as the inlet boundary condition for the DEM-

CFD simulations.

Moreover, in order to ensure that the flow distribution at the distributor

obtained via the comparison of the calculated and measured particle velocity is

reasonable, a CFD model for single-phase airflow was developed. This CFD

model gives the fractions of the fluidization airflow passing though the central

region of the distributor and through the outer annulus region. This single-phase


CFD model includes the air supply chamber, the bowl-shaped distributor with

orifices, the wire mesh screen and the fluidized bed. More details on this single-

phase model are provided in the Appendix.

5.2.2. Model validation

In order to evaluate the DEM-CFD model for particle motion, a simulation for

the base case was carried out. Particle motion was investigated in detail with

respect to the particle velocity, and the CTD and the RTD of particles in different

regions.

In Figure 5.4, the particle velocity field given by the DEM-CFD simulation and

measured using the PEPT experiment shows good qualitative agreement.

Quantitatively, as shown in Figure 5.5 and Figure 5.6, there is also good

agreement between the calculated and measured CTD and between the

calculated and measured mean residence times of particles in different regions.

Figure 5.4 The particle velocity field at a vertical cross-section through the center of

the bed (a) simulated using DEM-CFD and (b) measured using PEPT (VMD 1749 μm,

batch size 200 g, fluidization airflow rate 73.3 m3/h and atomization airflow rate

3.50 m3/h).


Figure 5.5 The cycle time distribution (for cycles shorter than 25 s) calculated in the

DEM-CFD simulation and measured in the PEPT experiment (VMD 1749 μm, batch size

200 g, fluidization airflow rate 73.3 m3/h and atomization airflow rate 3.50 m3/h).

Figure 5.6 The mean residence times of particles in different regions (VMD 1749 μm,

batch size 200 g, fluidization airflow rate 73.3 m3/h and atomization airflow rate 3.50

m3/h).

0

5

10

15

0 5 10 15 20 25

Perc

enta

ge o

f part

icle

cycle

s (

%)

Cycle time (s)

DEM-CFD

PEPT

1.27

0.27 0.25

5.61

1.33

0.27 0.30

6.61

0

2

4

6

8

The Wurstertube

The fountainregion

The downbedregion

The horizontaltransport

region

Resid

en

ce t

ime (

s)

Specific regions

DEM

PEPT


Figure 5.7 The mean cycle time and RSD for different batch sizes (VMD 1749 μm,

fluidization airflow rate 73.3 m3/h and atomization airflow rate 3.50 m3/h).

Furthermore, as an interesting observation in the PEPT experiments, the cycle

time, as well as the relative standard deviation (RSD), decreases when the batch

sizes increases. This effect is beneficial to process operation since it suggests that

larger batches can be coated with a more uniform coating thickness (i.e. better

quality) faster. This phenomenon was further investigated using the DEM-CFD

simulations, which show a similar trend, as illustrated in Figure 5.7. This effect

can be attributed to the fact that for the increased batch size, particles accelerate

faster in the Wurster tube, which gives fewer particle recirculations and less

cycle time.

Although the absolute values of cycle time and the RSD for each batch size still

exhibit slight differences between the experiments and the simulations, it can be

concluded that the DEM-CFD model for particle motion works for the present

system. A possible step forward for improvement would be to incorporate the

actual PSD instead of the VMD in the DEM-CFD simulations or to consider the

shape of the particles.

7.37

4.23

2.96

8.50

6.21

4.86

54

38

19

63 66

53

0

50

100

150

200

0

2

4

6

8

10

0 100 200 300 400 500 600 700

RS

D (

%)

Mean

cycle

tim

e (

s)

Batch size (g)

Mean cycle time by DEM-CFD

Mean cycle time by PEPT

RSD by DEM-CFD

RSD by PEPT


5.2.3. Evaluation of drag model (Paper III)

Figure 5.8 The cycle time distribution of particle cycles shorter than 25 s (VMD

1749 μm, batch size 200 g, fluidization airflow rate 73.3 m3/h and atomization airflow

rate 3.50 m3/h).

Figure 5.9 The residence times of particles in different regions for the base case (VMD

1749 μm, batch size 200 g, fluidization airflow rate 73.3 m3/h and atomization airflow

rate 3.50 m3/h).

0

5

10

15

20

25

30

35

0 5 10 15 20 25

Pe

rce

nta

ge

of p

art

icle

cycle

s (

%)

Cycle time (s)

PEPT

Gidaspow

Min

HKL

Beetstra

Tang

0

2

4

6

8

0

0.5

1

1.5

2

The Wurstertube

The fountainregion

The downbedregion

The horizontaltransport

region

Resid

en

ce

tim

e in t

he

ho

rizo

nta

l tr

an

sp

ort

regio

n (

s)

Re

sid

en

ce t

ime

s (

s)

Specified regions (-)

PEPT Gidaspow Min HKL Beetstra Tang


In order to determine if it is possible to improve the DEM-CFD model by refining

the model for the interphase momentum transfer, four other drag models were

examined. Figure 5.8 and Figure 5.9 show the CTD and the residence times of

particles in different regions, respectively, from the DEM-CFD simulations using

different drag models. When examining the particle velocity at different heights,

only a slightly higher particle velocity in the Wurster tube was found for the

HKL and Beetstra drag models. However, this slight difference in the particle

velocity in the Wurster tube can result in a distinct difference in particle

recirculations, i.e. due to particles sneaking out from below the lower edge of the

Wurster tube.

A different number of particle recirculations certainly cause differences in the

residence time of particles in the horizontal transport region and in the cycle

time. Therefore, the predicted coating quality can depend to a great extent on the

choice of drag models. Among these tested drag models, the one recently

proposed by Tang et al. (2015) showed good agreement with the PEPT data

(without any increase in computational cost) and should be considered in future

work. The investigation here again emphasizes the significance of the detailed

particle motion in the Wurster tube and suggests that care must be taken when

selecting the drag model for similar fluidized beds.

5.3. A predictive model for growth of pellets (Paper II

and Paper IV)

In order to understand the particle motion near the spray nozzle and how it can

affect the coating process, a spray zone in the shape of a solid cone was defined

as shown in Figure 5.10. This spray zone is mainly conceptual and is similar to

5.3. A predictive model for growth of pellets (Paper II and Paper IV) 45

the one in the study by Fries et al. (2011). The height of the spray zone, 𝐿, is

assumed to be 45 mm and the spray half-angle, 𝜃, is assumed to be 30 degrees.

Figure 5.10 A schematic of the spray zone.

In order to shed light on the underpinning mechanisms of how pellets with a

certain PSD grow differently during coating, mixtures of small and large pellets

were studied. Parameters such as the cycle time, the residence time of particles

in the spray zone and the entrance distance into the spray zone, i.e. the distance

between the spray nozzle and the location where particles move into the spray

zone (see Figure 5.10), were investigated. Among these parameters, the

residence time of particles in the spray zone and the entrance distance were not

available in the PEPT experiments, due to the poor data quality in this region. It

was, however, possible to obtain all these parameters in the DEM-CFD

simulations, but nonetheless time-consuming for the cycle time. Therefore, the

cycle time was taken from the PEPT experiments while the RTD in the spray

zone and the entrance distance into the spray zone were obtained from the DEM-

CFD simulations.

In the PEPT experiments, it was found that large pellets have a longer mean

cycle time than small pellets (c.f. Figure 5.11), which is equivalent to large

pellets traveling less frequently through the spray zone. In addition, a smaller

difference in cycle time was observed for the mixture with a larger fraction of


large pellets. On the other hand, the results obtained from the DEM-CFD

simulations show that large pellets spend a longer time in the spray zone and

move closer to the spray nozzle (c.f. Figure 5.12 and Figure 5.13). As the fraction

of large pellets in mixtures increases, the difference in the residence time in the

spray zone between the different mixtures increases accordingly.

Based on this data, the effects of the residence time in the spray zone, the

entrance distance into the spray zone, the cycle time and the particle diameter

on the coating thickness were explored with the predictive model described below.

The total amount of coating per unit time deposited on the surface of a particle,

𝑚𝑓𝑖𝑙𝑚, can be expected to be proportional to the spray rate, �̇�, the time spent in

the spray zone, 𝑡𝑠𝑝𝑟𝑎𝑦, and the number of coating cycles, 𝑁𝑐𝑦𝑐𝑙𝑒. The probability

that a spray droplet is deposited onto a particle can be expected to be

proportional to the cross-sectional area of the particle and to the droplet flux,

which decreases quadratically with the distance from the spray nozzle for a solid-

cone spray zone. Thus this model can be written as

𝑑𝑚𝑓𝑖𝑙𝑚

𝑑𝑡= 𝐾(𝑟𝑒𝑛𝑡𝑟𝑎𝑛𝑐𝑒)�̇� (

𝑡𝑠𝑝𝑟𝑎𝑦

𝑡𝑐𝑦𝑐𝑙𝑒)(

𝜋𝑑𝑝2 4⁄

𝑟𝑒𝑛𝑡𝑟𝑎𝑛𝑐𝑒2) (5.1)

where 𝑡𝑐𝑦𝑐𝑙𝑒~1/𝑁𝑐𝑦𝑐𝑙𝑒 is the cycle time and 𝐾 represents other factors that may

affect the amount of coating solution that a particle receives such as the

shielding effect. Since the shielding effect is predominantly entrance distance

dependent, 𝐾 should be a function of the entrance distance, i.e. 𝐾~1/𝑟𝑒𝑛𝑡𝑟𝑎𝑛𝑐𝑒 .

The rate of increase in the coating thickness can then be written as

𝑑𝑙𝑓𝑖𝑙𝑚

𝑑𝑡=

1

𝜌𝑓𝑖𝑙𝑚𝜋𝑑𝑝2

𝑑𝑚𝑓𝑖𝑙𝑚

𝑑𝑡 (5.2)

where 𝜌𝑓𝑖𝑙𝑚 is the density of the coating film.


Figure 5.11 The measured cycle time of small and large particles in the spray zone for

different mixtures of small and large particles.

Figure 5.12 The simulated residence time of small and large particles in the spray zone

for different mixtures of small and large particles.

4.85 4.95

6.09

7.94

5.956.29

0

2

4

6

8

10

0 25 50 75 100

Mean

cycle

tim

e (

s)

Mass fraction of large particles in 200 g mixtures of 1779 µm and 2665 µm particles (%)

1749 µm tracer

2665 µm tracer

0.115 0.1110.119

0.170 0.173

0.186

0

0.05

0.1

0.15

0.2

0 25 50 75 100Mean

resid

en

ce t

ime i

n t

he s

pra

y z

on

e (

s)

Percentage of large particles in 200 g mixtures of 1749 µm and 2665 µm particles (%)

1749 µm

2665 µm


Figure 5.13 The simulated entrance distance into the spray zone of small and large

particles for different mixtures of small and large particles.

Figure 5.14 The predicted relative rate of increase in the coating thickness between the

large and small particles for different mixtures of small and large particles.

26.1

26.5

25.9

24.6

24.224.4

24

25

26

27

0 25 50 75 100

En

tran

ce d

ista

nce i

nto

th

e s

pra

y z

on

e

(mm

)

Percentage of large particles in 200g mixtures of 1749 µm and 2665 µm particles (%)

1749 µm

2665 µm

1.08

1.691.81

0

0.5

1

1.5

2

0 25 50 75 100

Th

e r

ela

tiv

e r

ate

of

incre

ase i

n t

he

co

ati

ng

th

ickn

ess b

etw

een

th

e larg

e a

nd

sm

all p

art

icle

s (

-)

Percentage of large particles in 200g mixtures of 1749 µm and 2665 µm particles (%)


Using the above model, it is straightforward to predict the relative rate of

increase in the coating thickness between the large and small particles for

different mixtures, as shown in Figure 5.14. As seen in Figure 5.14, large

particles grow only slightly faster than small particles for the mixture with 25%

large particles. In this mixture, the longer residence time in the spray zone and

the closer entrance distance into the spray zone for larger particles are almost

counteracted by their less frequent passes though the spray zone. As the fraction

of large particles in the mixture increases, however, large particles show a

greater rate of increase in the coating thickness. This is well supported by the

fact that there is less difference in the cycle time (between large and small

particles) and a greater difference in the residence time in the spray zone

(between large and small particles).

Table 5.1 The size of the coated small and large particles for different mixtures.

Fraction of the

large particles

in the mixture

(%)

Class of particles

(-)

VMD

(µm)

Growth

(µm)

Relative rate of increase

in the coating thickness

between the large and

small particles

(-)

25 Small 1866 42

Large 2833 65 1.56

50 Small 1867 50

Large 2832 69 1.39

75 Small 1880 50

Large 2821 73 1.45

Although the predictive model is simple, the general understanding that it

provides can be expected to be applicable to other similar pieces of equipment.

Therefore, coating experiments using different equipment and a real two-fluid

atomization nozzle were performed to evaluate the performance of this model. In

these coating experiments, mixtures of the MCC pellets with the specifications

given in Table 4.2 were coated with EC/HPC polymer films. Table 5.1 shows the


size of the small and large pellets after coating for the different mixtures. It can

be seen that, during the same period of operation, large pellets had a greater

rate of increase in the coating thickness than small pellets: 1.56, 1.39 and 1.45

for mixtures with 25%, 50% and 75% large pellets. The magnitude of these

values is close to the prediction using the simple model and, additionally, clearly

show that large pellets grow faster than small pellets.

Despite qualitative agreement between the model predictions and the

experimental results, the trend of an increase in relative rate for the mixture

with more large particles was not confirmed in these experiments. Possible

refinements to include a more realistic representation of the equipment and to

include different model assumptions as well as micro-level phenomena such as

drying conditions in different regions are therefore suggested in Paper IV.

5.4. Particle drying

The experimental results, as can be seen in Paper V, are obtained for the mean

temperature and moisture content in air measured at different locations of the

fluidized bed. These results can be used for evaluating the performance of the

model for particle drying by comparing the simulated and measured values of

the temperature and moisture content in air at different locations. It is then

possible to access the temperature and moisture content for individual particles

at different times from the ongoing DEM-CFD simulations. In addition, the

drying rate of particles in different regions of the fluidized bed can be obtained.

6. Conclusions and outlook

In this thesis, with the overall objective of developing a mathematical model for

predicting the fluidized bed coating process, both experimental and modeling

efforts were focused on understanding the mechanisms that affect the coating

process. Overall, the following conclusions can be drawn.

A series of PEPT experiments were carried out to study particle motion in a

laboratory-scale Wurster fluidized bed. The measured particle trajectories were

used to identify particle cycles and to determine the CTD and the RTD of

particles in different regions of the fluidized bed. The effects of the partition gap,

the batch size, the Wurster tube length, and the fluidization and atomization

airflow rates on the CTD and the RTD of particles in different regions were

investigated. In this case, the experiments showed that particles spend most of

the cycle time in the horizontal transport region and, on average, approximately

12–29% of the time in the Wurster tube. It was observed that particles tend to

recirculate in the Wurster tube and sneak out from the bottom of the tube,

particularly if the partition gap is large, if the atomization flow rate is low or if

the Wurster tube is long. It was interesting to note that the cycle time decreased

and the CTD became narrower as the batch size increased.

Based on the parameters used in the PEPT experiments, coupled DEM-CFD

simulations were performed. The simulated particle trajectories were used to

determine the particle velocity field, the CTD and the residence times of particles

in different regions. The calculated results were then compared with the PEPT

data. The general characteristic particle motion was successfully captured in the

DEM-CFD simulations. The CTD and the residence times of particles in different

regions were found to correspond closely to the experimental results. A shorter

6. Conclusions and outlook 52

cycle time with a larger batch size was also predicted by the DEM-CFD

simulations.

In addition, in an effort to refine the model by examining different drag models,

it was found that satisfactory agreement could be obtained using the Gidaspow

and the Tang drag models. The agreement between the model predictions and

the experimental data leads to the conclusion that the DEM-CFD model for

particle motion is valid for the current system.

Using the validated DEM-CFD model, the detailed particle motion in a

simplified spray zone was studied for particles of different sizes. The results

showed that large particles spend a longer time in the spray zone and, on

average, move closer to the spray nozzle. Large particles obviously receive a

greater amount of coating solution due to their larger size, but the fact that they

travel closer to the spray nozzle than small particles provides evidence that large

particles can shield small particles from spray droplets. Both of these effects

suggested that large particles receive a greater amount of coating solution per

cycle than small particles. This, however, is partly counteracted by the fact that

large particles pass through the spray zone less frequently than small particles.

A simple conceptual model was then developed to predict the effect of the

residence time of particles in the spray zone, the cycle time, and the entrance

distance on the relative rate of increase in film thickness between large and

small particles. In addition, coating experiments were performed for evaluation

of the performance of this model. On comparing predicted and measured relative

rates of increase in coating thickness between large and small particles, a

qualitative agreement was obtained. It was confirmed that large particles grow

faster than small particles. Due to the simplicity of the model, several

refinements were suggested.

6. Conclusions and outlook 53

In a first model attempt, a model for particle drying was developed. Drying

experiments for parameter values similar to the ones used for the DEM-CFD

simulations were carried out to evaluate the performance of the model. Then it is

possible to access the temperature and moisture content for individual particles

at different times, as well as the drying rate of particles in different regions of

the fluidized bed, from the ongoing DEM-CFD simulations.

In all, this study demonstrates the power and potential of DEM as an important

tool for mechanistic studies of pellet coating in fluidized beds. Further steps

forward include the following. First, a better description of the atomization

airflow for real atomization nozzles is of interest. Although the simple model for

droplet deposition developed here was found to work quite well, it is nevertheless

of interest to gain a better understanding of the motion of particles and spray

droplets in the vicinity of the atomization nozzle. From a computational

perspective, this is challenging since it involves droplet breakup as well as effects

due to turbulence. Second, PSDs and possibly also the sphericity of particles may

be considered in further study. Third, it is also of interest to incorporate other

mechanisms into the model, such as agglomeration due to collisions between wet

particles and/or electronic charging. In addition, a reduction in the

computational cost would obviously also be very beneficial in order to handle

similar amounts of smaller particles or bigger batches of large particles. In the

end, it is of significance to tie all individual contributing mechanisms together in

order to further improve the predictive model that is in a multiscale nature.

References

Bommisetty, R. V. N., Joshi, D. S., & Kollati, V. R. (2013). Flow Loss in Screens: A Fresh Look at

Old Correlation. Journal of Mechanics Engineering and Automation, 3(1), 29-34.

Bruchmüller, J. (2011). Modelling the degradation of particles in fluidised beds. (Doctoral),

University of Southampton, PhD thesis. Retrieved from http://eprints.soton.ac.uk/333308/

Bruchmüller, J., Gu, S., Luo, K. H., & Van Wachem, B. G. M. (2010). Discrete element method for

multiscale modeling. Journal of Multiscale Modelling, 02(01n02), 147-162.

Cahyadi, C., Koh, J. J. S., Loh, Z. H., Chan, L. W., & Heng, P. W. S. (2012). A Feasibility Study

on Pellet Coating Using a High-Speed Quasi-continuous Coater. Aaps Pharmscitech,

13(4), 1276-1286.

Capecelatro, J., & Desjardins, O. (2013). An Euler–Lagrange strategy for simulating particle-

laden flows. Journal of Computational Physics, 238(0), 1-31.

Cheng, X. X., & Turton, R. (2000). The prediction of variability occurring in fluidized bed coating

equipment. I. The measurement of particle circulation rates in a bottom-spray fluidized

bed coater. Pharmaceutical Development and Technology, 5(3), 311-322.

Chiti, F. (2008). Lagrangian studies of turbulent mixing in a vessel agitated by a rushron turbine:

positron emission particle tracking (PEPT) and computational fluid dynamics (CFD).

(Ph.D.), University of Birmingham. Retrieved from http://etheses.bham.ac.uk/1607/

Chopra, R., Alderborn, G., Newton, J. M., & Podczeck, F. (2002). The influence of film coating on

pellet properties. Pharmaceutical Development and Technology, 7(1), 59-68.

Christensen, F. N., & Bertelsen, P. (1997). Qualitative description of the Wurster-based fluid-bed

coating process. Drug Development and Industrial Pharmacy, 23(5), 451-463.

Crowe, C. T., Schwarzkopf, J. D., Sommerfeld, M., & Tsuji, Y. (2012). Multiphase Flows with

Droplets and Particles, Second Edition. GB: CRC Press Inc.

Cundall, P. A., & Strack, O. D. (1979). A discrete numerical model for granular assemblies.

Geotechnique, 29(1), 47-65.

Darelius, A., Lennartsson, E., Rasmuson, A., Niklasson Björn, I., & Folestad, S. (2007).

Measurement of the velocity field and frictional properties of wet masses in a high shear

mixer. Chemical Engineering Science, 62(9), 2366-2374.

Deen, N. G., van Sint Annaland, M., van der Hoef, M. A., & Kuipers, J. A. M. (2007). Review of

discrete particle modeling of fluidized beds. Chemical Engineering Science, 62(1-2), 28-44.

Dintwa, E., Tijskens, E., & Ramon, H. (2007). On the accuracy of the Hertz model to describe the

normal contact of soft elastic spheres. Granular Matter, 10(3), 209-221.

http://eprints.soton.ac.uk/333308/

http://etheses.bham.ac.uk/1607/

References 56

Du, W., Bao, X., Xu, J., & Wei, W. (2006). Computational fluid dynamics (CFD) modeling of

spouted bed: Assessment of drag coefficient correlations. Chemical Engineering Science,

61(5), 1401-1420.

Ergun, S. (1952). Fluid flow through packed columns. Chemical Engineering Progress, 48(2), 89-

94.

Filho, R. S. C., Barrozo, M. A. S., Limaverde, J. R., & Ataide, C. H. (1998). Use of a spouted bed

in the fertilizer coating of soybean seeds. Drying Technology, 16(9-10), 2049-2064.

Fries, L., Antonyuk, S., Heinrich, S., & Palzer, S. (2011). DEM–CFD modeling of a fluidized bed

spray granulator. Chemical Engineering Science, 66(11), 2340-2355.

Gidaspow, D. (1994). Multiphase flow and fluidization: continuum and kinetic theory description.

Boston: Academic.

Green, D. W., & Perry, R. H. (2008). Perry's Chemical Engineers' Handbook (8th Edition):

McGraw-Hill.

Gunn, D. J. (1978). Transfer of heat or mass to particles in fixed and fluidised beds. International

Journal of Heat and Mass Transfer, 21(4), 467-476.

Hertz, H. (1882). Ueber die Berührung fester elastischer Körper. Journal für die reine und

angewandte Mathematik, 92, 156-171.

Hrenya, C. M., & Sinclair, J. L. (1997). Effects of particle-phase turbulence in gas-solid flows.

AIChE Journal, 43(4), 853-869.

Jajcevic, D., Siegmann, E., Radeke, C., & Khinast, J. G. (2013). Large-scale CFD–DEM

simulations of fluidized granular systems. Chemical Engineering Science, 98(0), 298-310.

Jenike, A. W. (1961). Gravity Flow of Bulk Solids. Bulletin No. 108, Utah State University.

Jono, K., Ichikawa, H., Miyamoto, M., & Fukumori, Y. (2000). A review of particulate design for

pharmaceutical powders and their production by spouted bed coating. Powder Technology,

113(3), 269-277.

Karlsson, S., Rasmuson, A., Niklasson Björn, I., & Schantz, S. (2011). Characterization and

mathematical modelling of single fluidised particle coating. Powder Technology, 207(1–3),

245-256.

Karlsson, S., Rasmuson, A., van Wachem, B., & Niklasson Björn, I. (2009). CFD modeling of the

Wurster bed coater. AIChE Journal, 55(10), 2578-2590.

Kearney, C. J., & Mooney, D. J. (2013). Macroscale delivery systems for molecular and cellular

payloads. Nat Mater, 12(11), 1004-1017.

Kona, R., Qu, H., Mattes, R., Jancsik, B., Fahmy, R. M., & Hoag, S. W. (2013). Application of in-

line near infrared spectroscopy and multivariate batch modeling for process monitoring

in fluid bed granulation. International Journal of Pharmaceutics, 452(1–2), 63-72.

Li, L., Rasmuson, A., Ingram, A., Johansson, M., Remmelgas, J., von Corswant, C., & Folestad, S.

(2015). PEPT study of particle cycle and residence time distributions in a Wurster fluid

bed. AIChE Journal, 61(3), 756-768.

References 57

Li, L., Remmelgas, J., van Wachem, B. G. M., von Corswant, C., Johansson, M., Folestad, S., &

Rasmuson, A. (2015). Residence time distributions of different size particles in the spray

zone of a Wurster fluid bed studied using DEM-CFD. Powder Technology, 280(0), 124-134.

Li, L., Remmelgas, J., Wachem, B. G. M. v., Corswant, C. v., Folestad, S., Johansson, M., &

Rasmuson, A. (2016). Effect of Drag Models on Residence Time Distributions of Particles

in a Wurster Fluidized Bed: a DEM-CFD Study. KONA Powder and Particle Journal,

10.14356/kona.2016008.

Mallouppas, G., & van Wachem, B. (2013). Large Eddy Simulations of turbulent particle-laden

channel flow. International Journal of Multiphase Flow, 54(0), 65-75.

Mann, V., & Crosby, E. (1975). Cycle time distribution measurements in spouted beds. Canadian

Journal of Chemical Engineering, 53(5), 579-581.

Marucci, M., Holmgren, A., Carlsson, H., Jarke, A., Johansson, M., & von Corswant, C. (2012).

Non-Uniformity of Pellets Coating, Effect on the Dose Release Profile and How to

Improve the Coating Process by Reducing the Electrostatic Charging of the Pellets.

Chemical and Biochemical Engineering Quarterly, 26(4), 379-384.

Pandey, P., & Bindra, D. S. (2013). Real-time monitoring of thermodynamic microenvironment in

a pan coater. Journal of Pharmaceutical Sciences, 102(2), 336-340.

Parker, D. J., Allen, D. A., Benton, D. M., Fowles, P., McNeil, P. A., Tan, M., & Beynon, T. D.

(1997). Developments in particle tracking using the Birmingham Positron Camera.

Nuclear Instruments and Methods in Physics Research Section A: Accelerators,

Spectrometers, Detectors and Associated Equipment, 392(1–3), 421-426.

Parker, D. J., Broadbent, C. J., Fowles, P., Hawkesworth, M. R., & McNeil, P. (1993). Positron

emission particle tracking - a technique for studying flow within engineering equipment.

Nuclear Instruments and Methods in Physics Research Section A: Accelerators,

Spectrometers, Detectors and Associated Equipment, 326(3), 592-607.

Parker, D. J., Forster, R. N., Fowles, P., & Takhar, P. S. (2002). Positron emission particle

tracking using the new Birmingham positron camera. Nuclear Instruments and Methods

in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated

Equipment, 477(1–3), 540-545.

Paulo Filho, M., Rocha, S. C. S., & Lisboa, A. C. L. (2006). Modeling and experimental analysis of

polydispersed particles coating in spouted bed. Chemical Engineering and Processing:

Process Intensification, 45(11), 965-972.

Rhie, C. M., & Chow, W. L. (1983). Numerical study of the turbulent flow past an airfoil with

trailing edge separation. AIAA Journal, 21(11), 1525-1532.

Seville, J. (2010). A single particle view of fluidization. Paper presented at the The 13th

International Conference on Fluidization - New Paradigm in Fluidization Engineering.

Shelukar, S., Ho, J., Zega, J., Roland, E., Yeh, N., Quiram, D., Nole, A., Katdare, A., & Reynolds,

S. (2000). Identification and characterization of factors controlling tablet coating

uniformity in a Wurster coating process. Powder Technology, 110(1), 29-36.

Sudsakorn, K., & Turton, R. (2000). Nonuniformity of particle coating on a size distribution of

particles in a fluidized bed coater. Powder Technology, 110(1–2), 37-43.

References 58

Tang, Y., Peters, E. A. J. F., Kuipers, J. A. M., Kriebitzsch, S. H. L., & van der Hoef, M. A. (2015).

A new drag correlation from fully resolved simulations of flow past monodisperse static

arrays of spheres. AIChE Journal, 61(2), 688-698.

Teunou, E., & Poncelet, D. (2002). Batch and continuous fluid bed coating – review and state of

the art. Journal of Food Engineering, 53(4), 325-340.

Tsuji, Y., Kawaguchi, T., & Tanaka, T. (1993). Discrete particle simulation of two-dimensional

fluidized bed. Powder Technology, 77(1), 79–87.

Tsuji, Y., Tanaka, T., & Ishida, T. (1992). Lagrangian numerical simulation of plug flow of

cohesionless particles in a horizontal pipe. Powder Technology, 71(3), 239-250.

Turton, R. (2008). Challenges in the modeling and prediction of coating of pharmaceutical dosage

forms. Powder Technology, 181(2), 186-194.

Wen, C., & Yu, Y. (1966). Mechanics of Fluidization. Fluidization, Chemical Engineering Progress

Symposium Series 62, 100-111.

Wurster, D. E. (1959). Air-suspension technique of coating drug particles. A preliminary report.

Journal of the American Pharmaceutical Association, 48(8), 451-454.

Appendix: Flow distribution at the

distributor

In order to estimate the flow distribution at the distributor, which was

unfortunately not possible to measure in the PEPT experiments, a single-phase

CFD model was developed and is described as follows.

A 1/6th model of the Strea-1 fluid bed was considered including the wire mesh

screen and the distributor. The detailed geometry of the screen was not included.

Instead, the wire mesh screen was modeled as a region with a specified pressure

loss. In Ansys Fluent, the CFD code used for this purpose, this was done using a

porous zone model, as sketched in Figure A.1.

Figure A.1 A 1/6th model of the Strea-1 fluidized bed, including the wire mesh screen

as a porous zone.

The pressure drop through the porous zone, i.e. the wire mesh screen, was

modeled as an inertial resistance and calculated using (Green & Perry, 2008;

Bommisetty et al., 2013)

Appendix: Flow distribution at the distributor 60

𝑃𝐷 =1

2𝐵𝜌𝑔𝒖𝑔

𝟐 (A.1)

where 𝐵 is the inertial resistance coefficient,

𝐵 =𝐷

𝑙𝑝 (A.2)

In equation (A.2), lp is the thickness of the porous zone and 𝐷 is the pressure

loss coefficient, which can be expressed by

𝐷 =1

𝐶2(1 − 휀2

휀2) (A.3)

where ε is the porosity and 𝐶 is the so-called discharge coefficient. The porosity is

based on the ratio of the opening to the projected total area of the wire mesh

screen and typically ranges from 0.14 to 0.79; for the screen employed here, its

value was 0.36. The discharge coefficient is a function of the screen Reynolds

number, NRe, and may be written as

𝐶 = 0.1√𝑁𝑅𝑒 (A.4)

where 𝑁𝑅𝑒 = 𝐷𝑠𝑢𝑔𝜌𝑔 (𝛼𝜇𝑔)⁄ and 𝐷𝑠 is the aperture width. For the screen used in

the experiment, 𝐷𝑠 = 0.25 𝑚𝑚.

In the PEPT experiments, there was a certain spacing between the distributor

and the wire mesh screen, as shown in Figure A.1. This spacing varied from 3 to

5 mm. The result, as shown in Figure A.2, shows that a larger spacing between

the distributor and the wire mesh screen leads to less air to flow through the

central base of the distributor. For the spacing that corresponds to the PEPT

experiments, the airflow passing through the central base of the distributor was

found to be between 52.9% and 63.5% of the total fluidization airflow.

Appendix: Flow distribution at the distributor 61

Figure A.2 The percentage of total airflow passing through the central base of the

distributor after the porous zone, i.e. the wire mesh screen, with a spacing of 3 mm and 5

mm.

52.9

63.5

0

10

20

30

40

50

60

70

0 1 2 3 4 5 6

Airflo

w t

hro

ug

h th

e c

entr

al b

ase

a

fte

r th

e w

ire

me

sh s

cre

en

(%

)

Spacing (mm)

particle motion, coating and drying in wurster fluidized...

Documents